Federal learning privacy protection method based on homomorphic encryption and secret sharing

Abstract

The invention discloses a federal learning privacy protection method based on homomorphic encryption and secret sharing, wherein the method comprises two parts: a gradient protection method based on homomorphic encryption and a random number protection method based on secret sharing, and mainly aims at four stages: an initialization stage, a model training stage, a model aggregation stage and a model updating stage. The gradient protection method based on homomorphic encryption is adopted, gradient protection is achieved, meanwhile, aggregation of gradient ciphertexts can be completed, leakage of gradient privacy information is effectively prevented, and safe aggregation of the gradient can be achieved; through the random number protection method based on secret sharing, the gradient ciphertext is protected, and meanwhile, the random number protected by the gradient ciphertext is protected, so that collusion attacks among the aggregation server, the key generation server and the participants can be effectively prevented, and the security of gradient information in the interaction process of the aggregation server, the key generation server and the participants is further guaranteed.

Description

technical field [0001] The invention belongs to the field of federated learning security technology, and specifically relates to a federated learning privacy protection method based on homomorphic encryption and secret sharing. The method proposes two technologies, a gradient protection method based on homomorphic encryption and a random number protection method based on secret sharing. It can not only ensure the security of the gradient uploaded by the participants, but also prevent the gradient information of the participants from leaking to the key distributor. Background technique [0002] As a branch of artificial intelligence, deep learning requires a sufficient amount of data for training. However, due to privacy issues, this condition is often not met. For example, in the medical field, since medical data is often very sensitive, it usually contains personal privacy. Information, in the process of data sharing in multiple medical centers, will lead to the disclosure ...

AI Technical Summary

Problems solved by technology

[0004]Aiming at the privacy leakage problem existing in the existing federated learning, this invention proposes a federated learning privacy protection method based on homomorphic encryption and secret sharing, and the application is based on homomorphic The encrypted gradient protection method and the random number protection method based on secret sharing ensure the security of the training data of the participants and prevent the leakage of the private information of the training data of the participants

Images